© Neuronpedia 2026

Privacy & Terms Blog GitHub Slack Twitter Contact

Neuronpedia

Jacobian LensNEW

Natural Language

NEW Assistant AxisNEW Circuit TracerUPDATESteer SAE Evals Exports interp-engineNEWAPI Community Blog Privacy & Terms Contact

GPT2-SMALL · 8-TRES-DC · 1921 ｜ Neuronpedia

Home
GPT2-Small
Transcoders Residuals
8-TRES-DC
1921

INDEX

Explanations

the conjunction "and" in contexts related to governance and authority

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Embeds

Show PlotsShow ExplanationShow ActivationsShow Test FieldShow SteerShow Link

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

�

-0.79

 Hercules

-0.70

 comprises

-0.62

 Heroes

-0.62

-0.61

 Fear

-0.59

 Afric

-0.58

ⓘ

-0.57

 Torment

-0.57

Evil

-0.56

POSITIVE LOGITS

ÍÍ

0.77

nostic

0.69

cer

0.69

Provider

0.69

 Attorney

0.68

cil

0.67

ctors

0.66

olitical

0.65

lement

0.64

STATE

0.64

Activations Density 0.080%

No Known Activations